Vatlas vs VIES
Every VAT number looked up twice in the same second: once through Vatlas, once straight at VIES. Here are the results.
Loading the figures…
Loading the figures…
Country by country
Best first. Vatlas answers from a national register wherever one is open to it, and asks VIES everywhere else, which is why some countries pull away and others sit level. The error rate is the share of requests that got no answer at all; a register stating the number is not registered has answered. Completeness asks two questions of each request: can the company be named, and can it be located. It is counted over every request made, including the ones that came back with no record.
| Country | Response time | Vatlas vs VIES | Error rate | Completeness |
|---|---|---|---|---|
| Loading the figures… | ||||
How this was measured
-
Same number, same second
For each VAT number we make two requests a fraction of a second apart, and we alternate which one goes first. Comparing two test runs made on different days would let a bad day at VIES pass for a flaw in the product; asking both at the same moment removes that risk.
-
VIES exactly as you would get it
The VIES side is one request, asked once. Vatlas asks again when a country replies that it is busy, stops asking a country it has just seen fail, and can serve the last answer it holds. That is the work you would otherwise have to do yourself. Waiting and asking again takes longer than giving up, so on a country that is down Vatlas is the slower of the two, and the one that comes back with an answer. Both sides name the same requester and therefore make the same call to VIES, the one that comes back with a consultation number.
A question VIES does not answer
Speed and completeness can be measured. The complete report has no equivalent on the VIES side.
VIES gives you its own opinion and nothing else. Every source at once asks, in a single call, every source that country has: the national register we hold, the register's own live endpoint where it publishes one, and VIES. Each comes back with its verdict and its date, and a line says so when it did not answer. No source is left out because a faster one already replied.